Department of Health and Human Services Center for Drug Evaluation and Research Review of Epidemiologic Studies on Cardiovascular Risk with Selected NSAIDs

Department of Health and Human Services

Center for Drug Evaluation and Research

Review of Epidemiologic Studies on Cardiovascular Risk with Selected NSAIDs

David J. Graham, MD, MPH

Office of Drug SafetyCenter for Drug Evaluation and Research

February 17, 2005



Purpose and Methods

• To evaluate epidemiologic data from the published literature plus 2 currently unpublished studies evaluated by this reviewer

• Focus: studies providing estimates of risk of acute myocardial infarction in the setting of use of COX-2 selective NSAIDs or naproxen– PubMed search by specific NSAID, with cross-checking of

cited references



Comments on Estimation of Excess Cases of AMI

• Tomorrow, FDA will present its estimation of the number harmed by rofecoxib, modeling RCT survival curves

• Assumes “grace period” based on VIGOR & APPROVe– Unreliable due to extremely low statistical power– Based on total of small # MI events over duration of trial– Epi studies based on 3- to 50-fold more events: more power

– Based on epi data, rofecoxib risk begins early in therapy

– No “grace period”

• Patients enrolled in RCTs are generally healthier than “real-world”– Therefore, RCTs will underestimate true risk and population

impact because their background rate is lower



Overview of Epidemiologic Studies on COX-2 AMI Risk (1)

Source Ref Popln Person-Yrs Study Popln Design Group Size Observed Age

Ray Medicaid Cohort Nonuse 454 K 428 K 50-84

Graham HMO NCC Remote 1.4 M 2.3 M 18-84 Celecoxib Solomon Medicare CC Multiple 54 K - 65

Mamdani Ontario Cohort Nonuse 167 K 76 K 66

Kimmel Community CC Remote - - 40-75

Ingenix MCO Cohort Ibu/Diclo 424 K 177 K 40-64

Medi-Cal Medicaid NCC Remote 651 K 2.4 M 18-84




Number of Cases Case Rofecoxib CelecoxibStudy Defn Cases All doses 25 mg >25 mg All doses

Ray +SCD 5,316 68 55 13 74

Graham +SCD 8,143 68 58 10 126

Solomon HAMI 10,895 225 202 23 425

Mamdani HAMI 701 58 - - 75

Kimmel HAMI* 1,718 27 25 2 18

Ingenix +SCD 628* 124* 83* 9* 139*

Medi-Cal HAMI 15,343 1,117 960 157 1,862




SpecialStudy Aspirin Smoking Features Limitations

Ray No No New-user analysis Low rofecoxib use; no med recs possible dose misclassification

Graham No* No* Inception cohort Low rofecoxib use; *Survey of controls

Solomon No No Duration analysis No SCD; multiple comparisons; Beneficiary survey no med recs; possible dose misclassification

Mamdani No No - Low rofecoxib use; prevalence cohort; excluded < 30 d users; no SCD; no dose/duration analysis; no med recs

Kimmel Yes Yes Direct interview Low rofecoxib use, very low high dose use; 55% case/50% control participation; no fatal AMI or SCD; self-report-?recall bias

Ingenix No No NDI search Identified 1798 cases; included only 628; New-user analysis no non-user reference; low high-dose use; Med record review possible dose misclassification

Medi-Cal Yes No Inception cohort New database for research purposes; no med recs; possible dose misclassification



California Medicaid

• Strengths

– Large sample size – over 7 million persons per year

– OTC aspirin data

– No censoring at age 65 (dual coverage with Medicare)

– Matching with multiple cause-of-death data

– Long durations of follow-up with low drop-out rates

– Sicker population than private-payors, so easier to detect drug safety

signals

• Limitations

– No access to medical records (HIPAA)

– Very complicated data – difficult to understand and analyze. Therefore, not

used often for drug safety research.



Unmeasured CV Risk Factors and NSAID Use

Graham Celecoxib Rofecoxib Naproxen Remote Aspirin 19% 23% 28% 24%

Smoking 9% 7% 11% 11%OTC NSAIDs 15% 14% 12% 13%

Solomon Celecoxib Rofecoxib NSAIDsBMI 27.5 27.2 27.7Aspirin 8.2% 11.5% 10.2%Smoking 8.7% 7.0% 9.8%College+ 29.6% 31.8% 26.5%Income Same Same Lower

Kimmel Celecoxib Rofecoxib NSAIDs RemoteBMI 29.7 28.0 27.7 27.2Aspirin 27.6% 32.1% 21.7% 28.8%Smoking Current 17.2% 6.5% 19.9% 21.2% Past 43.7% 42.9% 31.8% 32.0%Physical activity 7.0 7.2 7.4 7.4



Risk of AMI with Rofecoxib

Study Ref All doses 25 mg >25 mg

Ray Non - 1.02 (0.76-1.37) 1.93 (1.09-3.43)

Graham Rem 1.34 (0.98-1.82) 1.23 (0.89-1.71) 3.00 (1.09-8.31)

Solomon Rem 1.14 (1.00-1.31) - -

Mamdani Non 1.00 (0.80-1.40) - -

Kimmel Rem 1.16 (0.70-1.93) - -

Ingenix Active 1.41 (1.07-1.84) 1.54 (1.15-2.04) 0.81 (0.41-1.60)*

Medi-Cal Rem 1.32 (1.22-1.42) 1.29 (1.19-1.40) 1.56 (1.28-1.90)



Preliminary Results: Medi-Cal Study Dose Response for AMI Risk with Rofecoxib

0

1

2

3

4

5

6

7

8

Rofecoxib <= 12.5mg

Rofecoxib 12.5 -25 mg

Rofecoxib 25 - 50mg

Rofecoxib > 50 mg

Od

ds

Rat

io

1.161.31

1.54

2.40

Singh et al.



Risk of AMI for Rofecoxib vs Celecoxib

RofecoxibStudy All doses 25 mg >25 mg

Ray - - 2.20 (1.17-4.10)

Graham 1.59 (1.10-2.32) 1.47 (0.99-2.17) 3.58 (1.27-10.11)

Solomon 1.24 (1.05-1.46) 1.21 (1.01-1.44) 1.70 (1.07-2.71)

Kimmel 2.72 (1.24-5.95) - -

Medi-Cal 1.22 (1.11-1.33) - -



Individual Excess Risk of AMI or SCD per Year from Rofecoxib Use for an Average 65-74 Year Old US Man, based on

Epidemiolgic Data

Based on Based onpoint estimate 95% upper bound

25 mg

Ray 1/2500 1/135Graham 1/217 1/70Ingenix 1/93 1/48Medi-Cal 1/172 1/125

>25 mg

Ray 1/54 1/21Graham 1/25 1/7Medi-Cal 1/89 1/56



Excess Population Risk of AMI or SCD in 1 Million US Men 65-74 Years Old Treated with Rofecoxib per Year, Based on

Epidemiolgic Data

Based on Based onpoint estimate upper 95% bound

25 mg

Ray 400 7,400Graham 4,600 14,200Ingenix 10,800 20,800Medi-Cal 5,800 8,000

>25 mg

Ray 18,600 48,000Graham 40,000 146,200Medi-Cal 11,200 18,000



AMI Risk with Rofecoxib and Duration of Use

Graham et al. 50% 75% 95% 25 mg <2 m 5 m 13 m >25 mg <3 m 6 m 9 m

Solomon et al. Days 1-90 25 mg 1.37 (1.15-1.63) >25 mg 1.38 (0.80-2.37)

Kimmel et al. 25/27 cases 25 mg 102/105 patients 12 m

Days 1-30Solomon et al. 1.43 (1.12-1.83) Ingenix 1.51 (0.98-2.34)



Risk of AMI with Celecoxib

0.2

0.4

0.6

0.8

1

1.2

1.4

Ray Graham Solomon Mamdani Kimmel

Rel

ati

ve

Ris

k



Risk of AMI with Celecoxib – the effect of dose

0.6

0.8

1

1.2

1.4

1.6

1.8

Ingenix All Ingenix 200 mg Ingenix 400 mg MediCal All MediCal <=200mg

MediCal >200mg

Rel

ati

ve

Ris

k



Preliminary DataRisk of AMI with Valdecoxib

# cases OR (95% CI)

Medi-Cal 54 0.99 (0.72-1.37)

Mostly 10 and 20 mg. Medi-Cal only reimburses 10 mg tabs



Overview of Epidemiologic Studies on Naproxen AMI Risk (1)

Source Ref Person-Yrs Study Popln Design Group Observed Age

Ray Medicaid Cohort Nonuse 428 K 50-84

Graham HMO NCC Remote 2.3 M 18-84

G Rodriguez GPRD NCC Nonuse - 50-84

Rahme Quebec NCC Other NSAIDs - 65

Mamdani Ontario Cohort Nonuse 76 K 66

Schlienger GPRD NCC Nonuse - 18-75

Kimmel Community CC Remote - 40-75

Solomon Medicaid/Medicare CC Remote - -

Watson GPRD NCC Nonuse - 40-79

Ingenix MCO Cohort Ibu/Diclo 177 K 40-64

Medi-Cal Medicaid NCC Remote 52 K 18-84




Case NaproxenStudy Defn Aspirin Smoking Cases

Ray +SCD No No 201

Graham +SCD No* No* 367

G Rodriguez +SCD No Yes 49

Rahme HAMI Yes* No 397

Mamdani HAMI No No 15

Schlienger HAMI No Yes 19

Kimmel HAMI* Yes Yes ?

Solomon HAMI No No ?

Watson Composite No* Yes 26*

Ingenix +SCD No No 179*

Medi-Cal HAMI Yes No 368




Study Limitations

Ray Poisson assumption of constant hazard

G Rodriguez Definition of current exposure included potentially unexposed time (Rx ended within 30d of index date)

Rahme Excluded past AMI; relied on current exposure to another NSAID as reference; no external reference

Schlienger Small # events; excluded patients with underlying CV disease

Kimmel 2o or 3o analysis; small # events

Solomon Misclassified exposure (any exposure in past 6 mos); excluded patients with CV risk; adjustment based on diagnoses rather than Rxs

Watson Small # events; excluded prior CV disease & Rxs; composite outcome (AMI, CVA, SAH, SDH); failed to adjust, or poorly adjusted for CV risk



Risk of AMI with Naproxen

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

Ray Graham Rodriguez Mamdani Schlienger Ingenix MediCal

Rel

ati

ve

Ris

k



A Closer Look at 4 “Positive” Naproxen Studies

Solomon

Any use Current use Recent use Remote use

0.84 (0.72-0.98) 0.86 (0.64-1.16) 0.84 (0.63-1.12) 0.76 (0.56-1.03)

Watson Age Sex Yr DM CVrisk Comorb Smoke DMARDs Steroids 0.61 (0.39-0.94) + + + + + + 0.57 (0.31-1.06) + + + + + + 0.53 (0.22-1.28) + + + + + +

Rahme

Current naproxen vs other NSAIDs: 0.79 (0.63-0.99)

Reanalyzed, current naproxen vs nonuse: 1.28 (1.10-1.49), p=.001

Kimmel

0.48 (0.32-0.73) Small #s; 50% participation rate; mixing of Rx & OTC use; “reverse” recall bias



Preliminary Data: Medi-Cal StudyOther NSAIDs

# cases OR (95% CI)

Ibuprofen 719 1.11 (1.01-1.22)

Indomethacin 109 1.71 (1.35-2.17)

Meloxicam 81 1.37 (1.05-1.78)

Nabumetone 51 0.83 (0.60-1.14)

Sulindac 56 1.41 (1.01-1.96)

Non-coxib NSAIDs 2,006 1.12 (1.06-1.19)



Dose-Response Relationship of AMI risk with NSAIDs

0.5

1

1.5

2

2.5

Cel

eco

xib

<=

200

mg

Cel

eco

xib

> 2

00m

g

Dic

lofe

nac

<=

150

mg

Dic

lofe

nac

> 1

50m

g

Nap

roxe

n <

=10

00 m

g

Nap

roxe

n >

100

0m

g

Ro

feco

xib

<=

25

mg

Ro

feco

xib

> 2

5m

g

Od

ds

Rat

io



Preliminary Data: Medi-Cal StudyRisk of AMI Compared to Non-Coxib NSAIDs

OR (95% CI)

Celecoxib 0.97 (0.90-1.05)

Rofecoxib 1.18 (1.07-1.29)

Valdecoxib 0.88 (0.64-1.22)



Conclusions Regarding Risk of Acute MI:COX-2 Selective NSAIDs

• Celecoxib 200 mg: no apparent effect• > 200 mg: probable increased risk

• Rofecoxib 25 mg: probable increased risk• >25 mg: definite increased risk• Risk begins early in therapy, and is

apparent during days 1-30 of use

• Valdecoxib 20 mg: no apparent effect



Conclusions Regarding Risk of Acute MI:“Non-selective” NSAIDs

• As a class, non-coxib NSAIDs may increase risk

• Differences exist between non-coxib NSAIDs with respect to risk

• Naproxen is not cardio-protective



Open Questions

• Differential NSAID risk

• Dose response

• Duration effect

• Persistency of risk

• Actual benefit in the population

• CHF

• Stroke

Documents

Department of Health and Human Services Center for Drug Evaluation and Research Review of Epidemiologic Studies on Cardiovascular Risk with Selected NSAIDs